def test_classification_stats(self):
cfm = ConfusionMatrix(('a', 'not a'), ((20, 180), (10, 1820)))
self.assertEqual(20, cfm.tp)
self.assertEqual(180, cfm.fp)
self.assertEqual(10, cfm.fn)
self.assertEqual(1820, cfm.tn)
self.assertAlmostEqual(.67, cfm.sensitivity, 2)
self.assertAlmostEqual(.67, cfm.tpr, 2)
self.assertAlmostEqual(.91, cfm.specificity, 2)
self.assertAlmostEqual(.1, cfm.precision, 2)
self.assertAlmostEqual(.1, cfm.ppv, 2)
self.assertAlmostEqual(.91, cfm.accuracy, 2)
self.assertAlmostEqual(.91, cfm.acc, 2)
cfm = ConfusionMatrix(('a', 'b'), ((20, 5), (10, 15)))
self.assertAlmostEqual(.4, cfm.kappa(corrected=False), 2)
cfm = ConfusionMatrix(('a', 'b'), ((45, 15), (25, 15)))
self.assertAlmostEqual(.1304, cfm.kappa(corrected=False), 4)
cfm = ConfusionMatrix(('a', 'b'), ((25, 35), (5, 35)))
self.assertAlmostEqual(.2593, cfm.kappa(corrected=False), 4)
cfm = ConfusionMatrix(('a', 'b', 'c'),
((88, 10, 2), (14, 40, 6), (18, 10, 12)))
self.assertAlmostEqual(0.492, cfm.kappa(corrected=False), 3)
self.assertAlmostEqual(0.592, cfm.kappa(corrected=True), 3)
def test_manipulation(self):
cfm = ConfusionMatrix(('a', 'b', 'c'))
cfm['a', 'a'] += 10
cfm[0, 0] += 1
cfm['b', 'a'] = 2
cfm['b', 'b'] = 3
cfm[2, 1] = 6
cfm['c', 'c'] = 24
self.assertSequenceEqual((11, 0, 0, 2, 3, 0, 0, 6, 24), tuple(cfm))
self.assertSequenceEqual((11, 3, 24), cfm.diagonal)
self.assertEqual(46, cfm.total)
self.assertSequenceEqual((2, 3, 0), cfm[1])
self.assertSequenceEqual((2, 3, 0), cfm.row(1))
self.assertSequenceEqual((0, 3, 6), cfm.column(1))
def test_manipulation(self):
cfm = ConfusionMatrix(('a', 'b', 'c'))
cfm['a', 'a'] += 10
cfm[0, 0] += 1
cfm['b', 'a'] = 2
cfm['b', 'b'] = 3
cfm[2, 1] = 6
cfm['c', 'c'] = 24
self.assertSequenceEqual(
(11, 0, 0, 2, 3, 0, 0, 6, 24),
tuple(cfm))
self.assertSequenceEqual((11, 3, 24), cfm.diagonal)
self.assertEqual(46, cfm.total)
self.assertSequenceEqual((2, 3, 0), cfm[1])
self.assertSequenceEqual((2, 3, 0), cfm.row(1))
self.assertSequenceEqual((0, 3, 6), cfm.column(1))
def test_construction(self):
categories = ('a', 'b', 'c')
cfm = ConfusionMatrix(categories)
self.assertSequenceEqual(categories, cfm.categories)
self.assertEqual(3, cfm.size)
self.assertFalse(cfm.binary)
self.assertSequenceEqual([0] * 9, tuple(cfm))
self.assertRaisesRegex(ValueError, 'binary', cfm.get_binary_cell, 0, 0)
def test_binarization(self):
cfm = ConfusionMatrix(('a', 'b', 'c'))
cfm[0, 0] = 14
cfm[0, 1] = 3
cfm[1, 0] = 2
cfm[1, 1] = 36
cfm[1, 2] = 3
cfm[2, 0] = 1
cfm[2, 2] = 13
self.assertFalse(cfm.binary)
data = (('a', (14, 3, 3, 52)), ('b', (36, 5, 3, 28)), ('c', (13, 1, 3,
55)))
for category, expected in data:
bcfm = cfm.binarize(category)
self.assertTrue(bcfm.binary)
self.assertEqual(expected, tuple(bcfm))
def test_binarization(self):
cfm = ConfusionMatrix(('a', 'b', 'c'))
cfm[0, 0] = 14
cfm[0, 1] = 3
cfm[1, 0] = 2
cfm[1, 1] = 36
cfm[1, 2] = 3
cfm[2, 0] = 1
cfm[2, 2] = 13
self.assertFalse(cfm.binary)
data = (('a', (14, 3, 3, 52)),
('b', (36, 5, 3, 28)),
('c', (13, 1, 3, 55)))
for category, expected in data:
bcfm = cfm.binarize(category)
self.assertTrue(bcfm.binary)
self.assertEqual(expected, tuple(bcfm))
def test_classification_stats(self):
cfm = ConfusionMatrix(('a', 'not a'), ((20, 180), (10, 1820)))
self.assertEqual(20, cfm.tp)
self.assertEqual(180, cfm.fp)
self.assertEqual(10, cfm.fn)
self.assertEqual(1820, cfm.tn)
self.assertAlmostEqual(.67, cfm.sensitivity, 2)
self.assertAlmostEqual(.67, cfm.tpr, 2)
self.assertAlmostEqual(.91, cfm.specificity, 2)
self.assertAlmostEqual(.1, cfm.precision, 2)
self.assertAlmostEqual(.1, cfm.ppv, 2)
self.assertAlmostEqual(.91, cfm.accuracy, 2)
self.assertAlmostEqual(.91, cfm.acc, 2)
cfm = ConfusionMatrix(('a', 'b'), ((20, 5), (10, 15)))
self.assertAlmostEqual(.4, cfm.kappa(corrected=False), 2)
cfm = ConfusionMatrix(('a', 'b'), ((45, 15), (25, 15)))
self.assertAlmostEqual(.1304, cfm.kappa(corrected=False), 4)
cfm = ConfusionMatrix(('a', 'b'), ((25, 35), (5, 35)))
self.assertAlmostEqual(.2593, cfm.kappa(corrected=False), 4)
cfm = ConfusionMatrix(('a', 'b', 'c'),
((88, 10, 2), (14, 40, 6), (18, 10, 12)))
self.assertAlmostEqual(0.492, cfm.kappa(corrected=False), 3)
self.assertAlmostEqual(0.592, cfm.kappa(corrected=True), 3)
def test_construction_with_rows(self):
cfm = ConfusionMatrix(('a', 'b'), ((26, 3), (2, 72)))
self.assertSequenceEqual((26, 3, 2, 72), tuple(cfm))
self.assertEqual(26, cfm[0, 0])
self.assertEqual(2, cfm[1, 0])
